Legislative Staff and Representation in Congress ALEXANDER HERTEL-FERNANDEZ Columbia University MATTO MILDENBERGER University of California Santa Barbara LEAH C

American Political Science Review, Page 1 of 18 doi:10.1017/S0003055418000606 © American Political Science Association 2018 Legislative Staff and Representation in Congress ALEXANDER HERTEL-FERNANDEZ Columbia University MATTO MILDENBERGER University of California Santa Barbara LEAH C. STOKES University of California Santa Barbara egislative staff link Members of Congress and their constituents, theoretically facilitating democratic representation. Yet, little research has examined whether Congressional staff actually recognize the Lpreferences of their Members’ constituents. Using an original survey of senior U.S. Congressional staffers, we show that staff systematically mis-estimate constituent opinions. We then evaluate the sources of these misperceptions, using observational analyses and two survey experiments. Staffers who rely more heavily on conservative and business interest groups for policy information have more skewed perceptions of constituent opinion. Egocentric biases also shape staff perceptions. Our findings complicate assumptions that Congress represents constituent opinion, and help to explain why Congress often appears so unre- sponsive to ordinary citizens. We conclude that scholars should focus more closely on legislative aides as key actors in the policymaking process, both in the United States and across other advanced democracies.

INTRODUCTION electedofficials,thepublic,andinterestgroups.Aswewill document, despite their best efforts to estimate the epresentative democracy rests on the idea that fi preferences of their Members’ constituents, senior leg- elected of cials understand and act on their islative staff have very skewed perceptions of public ’ Rcitizens opinions. Yet research has struggled to attitudes. We also find that egocentric bias and interest demonstrate the link between representatives and their group contact, especially with conservative groups and constituents, especially in the United States. Instead, fi businesses, may drive some of the mismatch between there is growing evidence that elected of cials are more staffer perceptions and actual public opinion. responsive to interest groups and wealthier individuals We reach these conclusions using an original survey than to average citizens (Bartels 2008; Druckman and of senior legislative staffers in Congress merged with Jacobs 2015; Gilens 2012; Gilens and Page 2014) and mass public opinion data on five policies: gun control, that state politicians from both parties systematically carbon pollution restrictions, repeal of the Affordable overestimate the conservatism of their constituents 1 Care Act, infrastructure spending, and raising the (Broockman and Skovron 2018). Still, political scien- minimum wage. This approach allows us to examine tists do not have a clear picture of why legislators how well senior Congressional aides can characterize misperceive constituent preferences and respond so the public’s policy preferences. Across all five issues, we unequally to the mass public. In this paper, we examine find that staffers do not accurately identify their district one important and previously underappreciated mech- or state’s preferences and often overestimate their anism propagating inequalities in legislative responsive- constituents’ conservatism. We examine four explan- ’ ness: senior legislativestaffers perceptions, attitudes, and ations for the mismatch in staffers’ perceptions: electoral behaviors. Legislative staff act as a bridge between competitiveness; staffers’ personal policy preferences; staffers’ experience in Congress; and staffers’ inter- Alexander Hertel-Fernandez, Assistant Professor, School of Inter- actions with interest groups. We find that staffers’ per- national and Public Affairs, Columbia Universty, alexander.hertel@ sonal policy preferences and their interest group contact gmail.com. correlate most strongly with the opinion-representation Matto Mildenberger, Assistant Professor, Department of Political AU1 Science, University of California Santa Barbara. gap. Staffers whose personal opinions deviated from Leah Stokes, Assistant Professor, Department of Political Science, their constituents’ opinions were less accurate in their University of California Santa Barbara, [email protected]. estimates of district and state preferences. In addition, Authors listed alphabetically and contributed equally to the project staffers who reported greater contact with corporate and AU2 design and manuscript. ideologically conservative interest groups over liberal fi The authors thank Geoffrey Henderson for terri c research support and mass-based citizen groups—whether measured and are grateful to Raymond O’Mara III, Lee Drutman, and Kevin Kosar for aid in drafting and disseminating the survey. Thanks to Tim through staffers’ own reports or campaign contributions LaPira, Robert Shapiro, Kent Jennings, and participants at the UC to that staffer’s Member—were less likely to get their Santa Barbara Psychology, Environment and Public Policy (PEPP) constituents’ preferences right. Seminar and the Columbia Sustainable Development Workshop and We present results from two survey experiments the MPSA Political Institutions and Elite Behavior 4 mini-conference embedded in the legislative staffer survey that provide for feedback on earlier drafts. The Dirksen Congressional Center credible causal evidence for interest groups’ role in provided funding for the project through its Congressional Research fi explaining the opinion-representation gap. Using a list Grant program. Replication les are available at the American Political fi Science Review Dataverse: https://doi.org/10.7910/DVN/OWQNVF. experiment, we nd that about 45% of senior legislative staffers report having changed their opinion about Received: February 4, 2017; revised: March 27, 2018; accepted: 15 August 2018. First published online: nnn legislation after a group gave their Member a campaign contribution. In a second experiment, we show that 1 But see Enns 2015; Erikson 2015 for opposing perspectives. staffers are more likely to interpret correspondence Alexander Hertel-Fernandez, Matto Mildenberger, and Leah C. Stokes from businesses as being more representative of their still act as the heads of large “legislative enterprises,” constituents’ preferences than correspondence from with up to 18 full-time staff in US House offices and ordinary constituents. These findings provide evidence potentially dozens more in US Senate offices and that donor and interest group contact can shape how Congressional committees (cf. Salisbury and Shepsle staffers perceive policies, leading to constituent opinion 1981).3 mis-estimates. While some authors have argued that Congressional These results contribute to our understanding of staff ought to merely reflect Members’ preferences given Congress, legislative responsiveness, and political staffers’ career incentives for loyalty (DeGregorio 1988; inequality. Despite the fact that 80% of senior legislative Kingdon 1984, 1989), other research suggests that aides reported that constituent opinion was “extremely” staff exert a strong, independent effect on Member or “very” important in their recommendations to their behavior. Afterconducting extensive fieldwork in the bosses, most had only a limited ability to characterize 1970s, Malbin (1980)concludedthatwhilesome what their constituents actually preferred. Moreover, the Congressional staffers acted as effective surrogates prospect of more competitive elections—long thought to for their Members and constituents, in many other be an incentive for democratic representation—does not cases they were entrepreneurial, shaping legislation appear to spur greater congruence between senior on their own in meaningful ways. More recent staffers and their constituents. Instead, our observational analysis backs up this conclusion. Using a longi- and experimental evidence points to interest group tudinal dataset of staffers, Montgomery and Nyhan contact and egocentric bias as sources of the staffer- (2016)showthatMemberswhoexchangemore constituent representation gap. One reason why Con- senior staff behave more similarly than would oth- gress may seem so unresponsive to ordinary citizens is erwise be expected, even after taking into account that interest group lobbying and campaign contributions Member characteristics. skew staffers’ perceptions of their constituents. Since Given this evidence on staffers’ importance, this Congress, including staffers, hears muchmorefrequently paper poses three questions: To what extent do staffers from businesses, donors, and organized interest groups rely on their constituents’ preferences when crafting than individual citizens (Schlozman et al. 2012), staff may recommendations about policy for their Members? Are substitute interest groups’ and donors’ positions for staffers able to accurately assess those constituents’ constituent preferences. Our results highlight the need preferences? And, if not, why? These are similar for scholars of legislatures, lobbying, and public policy to questions to those long asked about Members of focusgreaterattentiononlegislativeaidesaskeyactorsin Congress (Erikson et al. 1975; Hedlund and Friesema the policymaking process, both in the United States and 1972; Miller and Stokes 1963); we instead place the focus other advanced democracies. on staffers (see also Broockman and Skovron 2018 for a similar approach to ours). Of course, staff may not possess the same incentives as politicians, and so it is worth reflecting on their LEGISLATIVE STAFF AND 4 POLITICAL REPRESENTATION relationships to politicians and constituents. Our the- oretical framework assumes that, as with lawmakers in Why focus on Congressional staff? Even as observers of general, staffers consider both career incentives and Congress note Members’ growing dependence on their personal policy preferences in their job. To advance on aides, few studies have explicitly examined staffers’ role Capitol Hill, we assume that staff face incentives to in the policymaking process. Yet, scholarly accounts of develop a reputation as competent and loyal aides to Congress give us good reason to think that staffers are an their Members. This entails helping Members to pass essential part of the legislative process (Kingdon 1984; legislation, service constituents, and engage in public Hall 1996; DeGregorio 1988; Hammond 1996; Malbin activities that permit Members to position themselves as 1980; Montgomery and Nyhan 2016; Romzek and Utter effective representatives. We assume that developing 1997; Fenno Jr. 1978). As Hall (1996) described it, “faced such a reputation is essential, even for staff who plan with the press of excessive obligations and the frequent to leave government. For instance, connections with prospect described of needing to be two places at once, Members are very valuable to aspiring lobbyists members have responded by relying increasingly on (Bertrand et al. 2014; McCrain, forthcoming). staff.” This power hasled someCongressionalscholars to For any given policy issue, staffers must gather and worry about staff serving as “unelected representatives,” synthesize information about the implications of the acting on their own to shape policy unmoored from electoral accountability (Malbin 1980). The presence of staff in Congress has changed dra- 3 While Representatives are allotted a maximum of 18 full-time, matically over time. Congressional staffing grew rapidly permanent staff members, US Senate office staffing depends on a from the 1950s through the late 1970s, reaching a peak in budget assigned to each office that takes into account the size of the the 1980s and followed later by a modest decline.2 Even Senator’s state and their distance from Washington, D.C. 4 with these reductions, however, Members of Congress Members of Congress have a strong incentive to select and retain staff who will be faithful “agents” to them; for instance, because they share a common set of priorities and preferences. While this selection and retention process complicates assessments of staffers’ effects on 2 See: https://www.brookings.edu/multi-chapter-report/vital-statistics- Members, it should not bias our efforts to understand how staffers on-congress/. perceive public opinion.

2 Legislative Staff and Representation in Congress various alternatives, as well as the stances of different staffs are more likely to mention resource-rich con- interests, including their Member’s constituents. Typi- stituencies, like businesses or interest groups, when cally, the senior-most policy staff in a Congressional recounting the actors who were important to them in office—Chiefs of Staff or Legislative Directors—will past policy debates. A final possibility is that staff take these various considerations and incorporate them accurately represent their constituents’ preferences, into a recommendation for their Member: for instance, but that any representation gaps are created elsewhere whether to vote for a bill, oppose a proposal, or release a in the legislative process. statement. Senior staffers have incentives to consider To gauge the extent to whichCongressionalstaffers district or state opinion when making recommendations are, or are not, representing their constituents’ totheir Membertohelptheirbossesappearresponsiveto preferences, we follow a long tradition in political their constituents (Hall 1996; Kingdon 1989). Gathering science and examine the difference between the constituent opinion, however, is not a costless endeavor perceptions that legislative staffers hold of their for staffers. As a result, not all staffers may be equally constituents’ opinions and their constituents’ actual able to judge the opinion of the citizens in their states and preferences (Broockman and Skovron 2018;Erikson districts across all policy issues. Staffers (especially in the et al. 1975;HedlundandFriesema1972;Millerand House) cannot regularly field representative polls of Stokes 1963). As in earlier studies,weoperationalize their constituents to figure out what those constituents political representation as the degree of congruence think. Instead, we assume that staffers primarily rely on between what constituentssaytheywantonsurveys alternative methods to gauge public opinion in their and what staffers think their constituents want on district: managing and aggregating constituent corre- those same policy issues. Our empirical strategy thus spondence; holding constituent meetings; hosting relies on merging an original survey of senior Con- townhalls; keeping in touch with community leaders; gressional staffers with survey data on Congressional following local media; and using interest group public constituencies’ opinions. opinion polls. It is worth acknowledging here that we are not Constituent preferences are not the only interests examining legislative staffers’ effect on specific policy that staffers rely upon when making recommendations choices or outcomes. Our evidence speaks to staffers’ to their Members, however. Staffers may also turn to abilities to correctly perceive their constituents’ pref- Congressional support agencies, think-tanks, mass- erences, not their ability to shape legislation. And, based membership groups, or trade associations. identifying a congruence between constituent opinion Especially if staffers do not have a good sense of what and legislative staffers’ perceptions of that opinion does their constituents want, staff may rely on interest groups not necessarily imply that constituent preferences are that claim to speak for other constituents, like busi- causally informing legislative staffers’ actions (indeed, nesses that claim to represent the economic interests of a see Jacobs and Shapiro 2000 for evidence to the contrary Member’s district or state. And staffers may also sub- in a related context). While recognizing these limi- stitute their own preferences and opinions when making tations, we note that staffers’ correct estimation of recommendations to their Members, or at least use their public opinion is a necessary though insufficient con- own preferences as a filter for understanding what their dition for citizens’ representation in Congress. If staffers constituents want. are not getting their constituents’ preferences right, Gaps between constituents’ opinions and staffers’ then it is unlikely that staffers’ inputs into the legislative perceptions could thus emerge in a variety of ways. process will reflect the majority preferences of the First, senior legislative staff might have idiosyncratic citizens in their districts and states.5 biases in their preferences for public policy that run against constituent attitudes. These biases might emerge from staffers’ ideologies, racial or ethnic THE 2016 LEGISLATIVE STAFFER SURVEY identities, personal finances, occupational experi- ences, or gender (e.g., Carnes 2013;Swers2002). If In August 2016, we fielded a survey instrument tar- staffers communicate their own preferences to their geting senior legislative staff in each Congressional bosses, rather than the public’spreferences,thiscould office. Typically, the target staffers’ job titles were either introduce a mismatch in representation. Second, ‘Chief of Staff’ or ‘Legislative Director’. We chose to staffers could have a different picture of what their survey these aides because these are the individuals constituents want based on the contact staffers have within each Congressional office responsible for with a subset of their constituents, like organized developing the legislative and political agenda for their interest groups, donors, or activists, perhaps because of their Member’selectoralpressuresortheirexpe- rience on the job. If a staffer only has contact with an especially vocal subset of their Member’sconstituency, they might well develop a distorted perception of their 5 We also acknowledge that we are focused on public preferences district or state (Stokes 2016). Indeed, this is often an related to specific policy issues, rather than the mass public’s more explicit strategy organizedinterestgroupsusewhen general ideological orientation. It may well be that individuals who espouse a “liberal” position on these specific issues would still express lobbying Members (Arnold 1990;Kingdon1984; a conservative political outlook (see canonically Free and Cantril Kollman 1998). Consistent with this strategy, Miler 1967). Such an extension of our study would be a fruitful line of future (2010)hasshownthatMembersofCongressandtheir research.

3 Alexander Hertel-Fernandez, Matto Mildenberger, and Leah C. Stokes

Member of Congress and for reporting directly to that participate in our study as a result of those policies. In Member (CMF 2011).6 Chiefs of Staff and Legislative other cases, however, staffers reported making an Directors are thus the individuals ultimately in charge of exception for our survey because of its confidential and reviewing the information an office receives from academic nature—two facts that we heavily emphasized ordinary constituents, interest groups, and other sour- in our correspondence. These policies would be con- ces, and then incorporating that information into rec- cerning if certain types of Congressional offices were ommendations for their Member. Thus, our survey more likely to implement them, or certain types of targeted the population that plays a crucial role in the staffers were more likely to follow these policies more policymaking process, connecting the preferences of strictly, causing our survey results to represent a biased constituents with Members of Congress. It may well be sample of staffers and offices. Reassuringly, however, the case that other staffers within a Congressional our sample of respondents is quite close to the overall office—like district directors—could more accurately population of senior legislative staffers and Congres- judge their constituents’ opinions. However, we view sional offices on a range of observable characteristics. Chiefs of Staff and Legislative Directors as the critical This provides evidence that certain types of offices or sample for our study’s research question because they staffers did not systematically answer our survey. are central in shaping Members’ policy decisions. Dis- Figure 1 compares our survey respondents to the trict staff’s perceptions of public opinion mean little if overall population of senior legislative staffers in they cannot be accurately conveyed through more Congress, contrasting differences in means for a variety senior staff to shape Member decision-making. of characteristics with 95% confidence intervals.9 Full The survey asked staffers a range of questions about balance tables are provided in Appendix C. In general, policy preferences, including staffers’ own preferences our sample closely resembles the overall population. and their beliefs about their constituents’ attitudes. It Still, there are some differences. The largest gap also asked staffers about where their policy information between our survey sample and the overall population comes from, including their reliance on organized of staffers was in partisanship: slightly over half of our interest groups. In addition, the survey included several sample (54%) came from Democratic offices, while in experiments. One list experiment measured staffers’ Congress as a whole only 43% of senior staffers work in willingness to change their beliefs on public policy as a Democratic offices. Nevertheless, we still have a suffi- result of campaign contributions from an organized cient number of Republican respondents to disaggregate interest group. Another experiment gauged how our analyses by party. Within both Democratic and staffers would respond to communications from dif- Republican respondents, moreover, the responding ferent constituents, including business and advocacy offices’ ideological orientations are quite similar to the groups. overall distribution of Congressional ideology, as To construct the senior Congressional staff pop- measured by standard first-dimension DW-NOMI- ulation, we used the Leadership Directories database to NATE ideal points (Carroll et al. 2015). identify Chiefs of Staff and Legislative Directors in every US House and US Senate office as of July 2016. In cases where staffers did not have those exact titles, we fi POLICY CONSIDERATIONS FOR identi ed other individuals who would serve as the top LEGISLATIVE STAFFERS policy staffers in an office. We sent an initial email invitation to these staffers to participate in our survey on Before assessing the relationship between staffer per- August 18, 2016. Two follow-up emails were sent to ceptions of their constituents and constituent opinion, staffers on August 24 and August 31.7 Our final survey we first consider the information sources that staffers sample includes 101 respondents from 91 offices, for a report relying on. This allows us to compare, in staffers’ total response rate of 9.6%, in line with other studies of own minds, the importance of constituent opinion Congressional staff.8 against other potential considerations. We asked A number of Congressional offices have policies staffers: “Think about the policy proposals you have against participating in surveys, and we received emails worked on during your time on the Hill. What shaped indicating that many staff would not be able to your thinking on whether your Member should support or oppose these policies? Indicate how important each of the following considerations was in shaping your 6 According to the guide developed by the Congressional Manage- advice to your Member on various policy proposals.” ment Foundation, the job description for Chief of Staff is “Top staff Figure 2 reports this item’s results by party. On person responsible for overall office functions; oversees staff and budget; advises Member on political matters; responsible for hiring, average, staffers reported constituent communication promoting, and terminating staff; establishes office policies and and attitudes were most important, and concerns about procedures”; the description for Legislative Director is “Establishes primary opponents were least important. Over 80% of legislative agenda; directs legislative staff; serves as resource person staffers view either constituent opinion or communi- for LAs [Legislative Assistants]; briefs Member on all legislative cation as extremely or very important in shaping their matters; reviews constituent mail.” advice to Members. These results fit well with past 7 See Appendix A for copies of recruitment materials. 8 polling of Congressional staffers by the Congressional For instance, our response rate is similar to the response rates the Congressional Management Foundation obtained (~15%) in its studies of Congressional offices (See: http://www.congressfoundation. org/publications/1048-managing-changes-in-budgets-and-benefits). 9 We clustered standard errors by Member office.

4 Legislative Staff and Representation in Congress

In follow-up, semistructured interviews with some of FIGURE 1. Balance Between Survey our survey respondents, we also probed staffers about Respondents and Overall Population of Staffers fi their efforts to process constituent communications and (with 95% Con dence Intervals) opinions.12 Our interviews indicated that to track constituent opinion offices generally log every phone call, e-mail, letter, and fax they receive into a database. These databases can be used to sort pieces of corre- spondence into batches relating to particular issue areas, such as health care or trade, and to report the number of contacts the office has received on each issue as well as the stances conveyed in the messages. When the American Health Care Act (the GOP bill to repeal the Affordable Care Act) was introduced in the House in 2017, for instance, staffers used this information to track how many contacts their office received in favor of the bill, and how many opposed it. While this process is remarkably consistent across offices, interviewees varied in the extent to which they reported that their office engages in additional methods of assessing constituents’ views. The other methods they cited include face-to-face meetings with constituents in their offices, polling, monitoring constituents’ com- ments on social media, and even door-to-door can- vassing by Members themselves. Several interviewees also mentioned town halls and other public events in the district or state, such as civic association meetings, as a useful means of keeping abreast of the issues that are Management Foundation, which has found that over most salient to constituents. 90% of surveyed staffers say that constituent contact Staffers in our interviews consistently indicated would be important in deciding an office’s stance on an correspondence tallies on a given bill (e.g., how many issue.10 In fact, Congressional offices appear to priori- people asked a Member to vote for or against it) are tize collecting and responding to constituent opinions so often incorporated into recommendations about much that the CMF has found that nearly half of sur- whether to vote for the bill. Similarly, interviewees veyed staffers said that their office had shifted resources reported that the information about which issues they away from other activities toward managing constituent are hearing about from their constituents plays a role in communications.11 This is understandable, given that their office’s decisions about which issues to focus on. the Internet makes it easier for constituents to com- For instance, at weekly meetings with the Member municate with their elected officials’ offices. Of course, where the office decides its priorities, staffers review these estimates are based on staffer self-reports. We which issues they have been hearing about most from might be concerned that, even in the context of a constituents during the past week. We will evaluate the confidential survey, staffers downplay their reliance on extent to which staffers actually perceive their con- certain information sources that are viewed as politi- stituents’ preferences and take correspondence from cally unpopular. To evaluate this potential bias, we also constituents as seriously as they indicate on this ques- asked staffers what considerations other offices used for tion in the sections that follow. persuasion during past policy debates. This question Returning to Figure 2, there were large differences by partially taps into revealed behavior, rather than simply party in the considerations staffers reported were rel- staffers’ own judgments. If another office was trying evant for their policy advice, aside from constituent hard to persuade a staffer and their Member about an opinion. The largest divide was on information from issue, they would likely pick considerations they knew unions: nearly half (49%) of Democratic staff reported would be taken more seriously by the target office. We that union information was extremely or very important provide full details on this question and results from this to their deliberations, while only 7% of Republican staff survey item in Appendix D. In general, there was a said the same. Another striking difference was the strong relationship between the types of considerations importance of information from party leaders (Curry and information sources described as personally 2015). Over half (55%) of Democratic staffers reported important by staffers and as frequently used by other that they found information from their party’s leader- offices for the purposes of persuasion. ship to be extremely or very important to them, while

12 Our research team conducted 30-minute interviews with 10 http://www.congressfoundation.org/storage/documents/ respondents from the survey from July through September 2017. CMF_Pubs/cwc-perceptions-of-citizen-advocacy.pdf. These interviews were conducted with 11 Democrats and seven 11 http://www.congressfoundation.org/storage/documents/ Republicans. Thirteen interview participants were Chiefs of State and CMF_Pubs/cwc-mail-operations.pdf. five were Legislative Directors.

5 Alexander Hertel-Fernandez, Matto Mildenberger, and Leah C. Stokes

FIGURE 2. Considerations Staffers Reported as Extremely or Very Important in Shaping Advice to Their Members, by Staffer Party

only one in five Republican staffers reported the same public opinion. In particular, these interviews fore- (21%). This divide may reflect the ongoing struggles closed the unlikely interpretation of our question as within the Congressional Republican caucus as GOP querying the narrow preferences of their Member’s leaders like Speaker Paul Ryan and Senate Majority political supporters. All our interviewees unanimously Leader Mitch McConnell attempt to bring together interpreted our question wording as referring to the more traditional, business-friendly Members with residents of their Member’s district or state. No staffer newer and more ideological Freedom Caucus and Tea interpreted it as referring to political supporters.13 Party members (Mann and Ornstein 2016). We selected the five policy domains included on the survey for several reasons: public opinion data avail- ability, issue salience, partisan polarization, and interest LEGISLATIVE STAFFERS’ PERCEPTIONS OF group involvement. With the exception of the climate THEIR CONSTITUENTS change item, the items came from the Cooperative Congressional Election Study (CCES), which offers Over three-quarters of both Democratic and Repub- sufficiently large sample sizes to estimate downscaled lican staffers reported in our survey that their top public opinion for states and Congressional districts. consideration when thinking about legislation was their The CCES uses respondents sampled from YouGov, a constituents’ opinion. But do staffers know what their well-respected online polling firm. The sample aims to constituents want? To answer this question, we com- be representative of the national adult population and is pare staffers’ estimates of public opinion in their district very large with around 50,000 respondents during or state to the actual opinion in that district or state. election years. We use data from the 2016 CCES to fi Speci cally, we ask respondents “Now, consider only estimate district- and state-specific attitudes using the people living in your Member of Congress’s con- multilevel regression and post-stratification (MRP), stituency. To the best of your knowledge, what per- detailed in Appendix F. Our climate question uses data centage of the people living in your Member’s previously down-scaled by Howe et al. (2015). In the constituency would agree with the following policy analyses that follow, we compare Senate staffer statements? The slider below goes from 0% (no one) to 100% (everyone). Just give your best guess.” 13 Respondents moved a slider bar to answer the question. Moreover, when asked whether constituents included people residing outside the district or state, some staffers mentioned that they As part of our follow-up interviews with survey received correspondence from people living outside the member’s respondents, we validated our interpretation of this jurisdiction, but that they were able to tell whether the contact came question as representing estimates of district-level from a resident of the district or state.

6 Legislative Staff and Representation in Congress perceptions with state-level opinion estimates, and this policy as engendering the opposition of a con- House staffers with district-level opinion estimates. For centrated set of corporate interests. Tighter gun regu- all five policy areas, we used essentially the same lations represent a similar issue where there is intense question wording in our survey as in these other corporate opposition from a concentrated set of inter- surveys.14 ests, in this case, the firearm and ammunition industry All five policy domains are highly salient, and and its advocacy groups, especially the National Rifle therefore are issues on which staffers have received at Association (Cook and Goss 2014). Raising the mini- least some constituent and interest group contact. This is mum wage is also an issue that business interests have especially true because Congress considered bills that opposed; but in this case, opposition is less concentrated would address all five proposals within the prior year to in particular industries and more widespread across the the survey.15 The fact that Congress had recently voted business community (Hacker and Pierson 2010). Health on measures related to these policies means that top reform, on the other hand, features a mixed set of legislative staffers ought to have considered, relatively corporate interests. Although health insurers and some recently, the positions of their constituents on these medical providers strongly opposed health reform issues. The five issues also exhibit varying degrees of legislation, including the ACA (Jacobs and Skocpol partisan polarization. In two cases—ACA repeal and 2015), these sectors face a different set of incentives now the minimum wage—the issue is highly polarized that the ACA is law. Repealing some or all of the ACA between the two parties. There is essentially no Con- would undermine these companies’ current business gressional Democratic support in favor of repeal and model, which now depends on enrolling Americans into virtually all Congressional Republicans were suppor- new regulated marketplaces and an expanded Medicaid tive of ending the health reform law, at least in principle system. At the same time, these industries would sup- before the Trump administration; while the reverse is port rolling back many individual provisions within the true for boosting the minimum wage. By contrast, both ACA. Lastly, infrastructure spending is an issue sup- parties have tended to be more divided internally ported by many segments of the business community as toward climate change and gun control: even though a useful investment in the American economy. The very Democrats have tended to be much more favorable conservative US Chamber of Commerce, typically toward measures to address climate change and gun opposed to most government intervention, backed the control than Republicans, there are more divisions on infrastructure spending bill considered by the 114th these issues within the parties. Some conservative Congress.16 Democrats, for instance, have stymied Congressional Figure 3 compares the relationship between Con- climate proposals and new gun control measures. Some gressional staffers’ perceptions of their constituents’ more moderate Republicans have occasionally sup- opinions and their constituents’ actual opinions in each ported legislation in these domains as well. And policy area.17 If staffers were able to accurately perceive infrastructure spending is generally viewed as a bipar- public opinion, the data should fall on the dashed 45 tisan issue historically supported by both parties as a degree line; when staffers are underestimating support, necessary investment in the economy. Indeed, an the dots fall to the left-hand side of the line, when they infrastructure spending bill was one of the few bipar- are overestimating support, they fall to the right. The tisan accomplishments of the otherwise gridlocked locally weighted (lowess) regression line shows that 114th Congress. staffers estimate public opinion across the full range, The five issues also implicate different constellations from 0% to 100% support. By contrast, public support of interest group involvement, which we anticipate may in each domain falls within a narrower range across help explain Congressional representation. On climate districts: from 50% to 82% for carbon regulation; 39% change, fossil fuel and other extractive industries to 69% for repealing the ACA; 73% to 95% for gun sale intensely oppose tighter regulations. These businesses regulations; 72% to 86% for infrastructure spending; have invested a considerable sum of money in campaign and 57% to 90% for boosting the minimum wage. While contributions, lobbying, and other political activities to there are modestly positive slopes across all five issues prevent government action to curb greenhouse gas when pooling across Democratic and Republican emissions (Mildenberger 2015). We might characterize

16 See, e.g.: https://www.uschamber.com/blog/its-time-raise-federal- 14 For more information on the CCES, see: https://cces.gov.harvard. gas-tax. edu/. Appendices B and E summarize the wording of our questions, as 17 We might be concernedthat these results are in some way an artifact well as the sources for the survey data we used to estimate district- and of sampling uncertainty in our MRP estimates. Fortunately, we have a state-specific attitudes using multilevel regression and post- strategy to directly evaluate and reject this possibility. The Howe et al. stratification (MRP). We detail our MRP models in Appendix F as (2015) data we rely on for our climate-based estimates involved well. external validation of the MRP estimates against independent local- 15 In December 2015, Congress voted on a measure to curb the EPA’s level polls to estimate the sampling uncertainty associated with the ability to regulate CO2 emissions. Congress voted on ACA repeal as model. In that research, Howe et al. report that congressional district recently as February, 2016, when the House voted to override Pres- estimates are within 67 percentage points of their external validation ident Obama’s veto of repeal legislation H.R. 3762. In June, 2016, dataset. Taking this estimate, we find that 66 staffers (72% of Congress voted on background check measures. In May 2015, respondents that answered this question) have a mismatch that is Democrats released a $12 an hour minimum wage proposal. And in greater than this uncertainty estimate. This should give us high con- December 2015, Congress voted on a major infrastructure spending fidence that staffer misperceptions are not a simple function of MRP bill. sampling uncertainty.

7 Alexander Hertel-Fernandez, Matto Mildenberger, and Leah C. Stokes

FIGURE 3. Comparing Staffer Perceptions of Constituent Preferences to True District-Level Preferences, by Issue Area

staffers, these correlations generally disappear once we across the five domains: the correlation between staffer consider within-party staffer misperceptions.18 Overall, estimates and constituent preferences was strongest for then, in none of the five areas are staffers estimating attitudes about raising the minimum wage, followed by their constituents’ preferences with any degree of rel- CO2 limits, repealing the ACA, gun sale checks, and ative or absolute accuracy.19 Staffer perceptions are far then infrastructure spending.20 more extreme than the public’s actual policy prefer- If staffers do not have accurate perceptions of publics in ences. Still, the strength of the relationship between their districts or states, do these perceptions instead reflect staffer estimates and constituent preferences varied the distribution of public attitudes among co-partisans? We can evaluate this possibility in the climate change issue area by leveraging data from a partisan MRP model of 18 Considering the 10 issue-party relationships between public opinion and staffers’ guesses, only two cases are statistically significant climate and energy attitudes (Mildenberger et al. 2017). In at conventional levels: Democratic staffers on the minimum wage and Figure 4,wecomparestaffers’ perceptions of their con- on repealing the ACA. The remaining eight slopes are all far from stituents’ preferences for carbon regulation with the statistical significance. fraction of the Democratic or Republican publics in 19 While we do not have the ability to estimatethe preferences of likely their elected official’s district supporting this position. voters—as opposed to all constituents—at the district level, we can We again find little relationship. Democratic and compare differences between the true national preferences among likely voters and all US citizens. Subsetting CCES data to individuals who report that they will definitely vote in a given election, we find negligibledifferences betweenthe averagepreferencesof likely voters 20 In Appendix Q, we also perform a robustness check on these results and all citizens. This suggests that the misperception gap is not a by redoing our analysis after dropping staffers who estimate that the function of staffers simply discounting the preferences of unlikely public in their district is over 90% or less than 10% in support of any voters from their estimates. Further MRP analysis that distinguished individual policy.We thus show that the surveyresults are robust when between validated voters and nonvoters would be fruitful to test this excluding staffers who offered potentially nonrealistic answers to possibility in even greater detail. these questions.

8 Legislative Staff and Representation in Congress

EXPLAINING LEGISLATIVE STAFFERS’ FIGURE 4. Staffer Perceptions of Constituent GAPS IN REPRESENTATION Preferences Against Distribution of Co-Partisan Constituent Preferences What could explain these strikingly large mismatches between staffer perceptions and constituent opinion? We examine four potential correlates for the repre- sentation gap: elections; staffers’ own preferences; staffer experience; and interest group contact. We summarize these hypotheses in Table 1.21 Standard models of representation predict that elections should drive responsiveness for politicians who want to win and hold office (Arnold 1990; Gilens 2012; Mayhew 2004). Therefore, we might expect that staffers whose Members were in tighter races at the time of the survey would be more likely to correctly assess constituent opinions because they want their Members to win reelection to retain their jobs. We assessed this hypothesis using a question on the survey that asked staffers to predict their Members’ reelection margin, if their Member was up for reelection in 2016. We received 75 valid responses, ranging from a predicted 2- to 100-percentage point victory, with a mean of 35 and a median of 23-percentage points. Notably, no staffer predicted a loss. Figure 5 plots the relationship between the staffer- constituent opinion mismatch and staffers’ election predictions. Looking across all policy domains, we see little consistent relationship between estimated race competitiveness and the accuracy of staffers’ percep- tions. Staffers for Members in tighter races were gen- Republican staffers’ estimates of the general publics’ erally no more likely to get their constituents’ preferences do not seem driven by the distribution of preferences right compared to staffers working for their co-partisan publics’ beliefs. If staffers’ perceptions Members in relatively uncontested races. The strongest of public opinion were structured by primary or general exception was for Republican staffers on gun policy. election partisan voters, we would have expected to see Here, we see the expected relationship: GOP staffers a stronger relationship in this figure. from safer districts and states were more likely to Our data also allow us to evaluate whether staffers are underestimate their constituents’ support for back- systematically over- or underestimating public attitudes. ground checks. Yet, even for staffers in highly com- Previous researchfinds thatboth liberaland conservative petitive races there was a large mismatch: staffers who state politicians systematically over-estimate their con- thought that their bosses would only win by two per- stituents’ conservatism (Broockman and Skovron 2018). centage points were estimated to be off of their con- We similarly find a conservative bias for four of our five stituents’ preferences by about 43-percentage points. issues. The vast majority of surveyed staffers (91%) Electoral competitiveness thus does not seem to offer underestimated their constituents’ support for back- much leverage in explaining the staffer-constituent ground checks on gun sales. There was a similar, though mismatch (see Appendix H for evidence that actual less pronounced, bias toward staffers underestimating race competitiveness does not predict representation, support for regulating carbon dioxide emissions—78% either). Next, we examined the role of staffers’ personal of staffers underestimated constituent support for CO2 limits in their districts and states. There was less of a clear opinions (Figure 6). For this hypothesis, we were pattern forstafferswhenestimating supportforrepealing interested in understanding whether egocentric bias could explain mismatches. Egocentric bias is a con- the ACA: 65% underestimated and 35% overestimated fi mass support. However, there were clear partisan divi- sistent nding in psychology that suggests individuals sions in staffer misperceptions. Every Democratic staffer use their own beliefs as a heuristic for estimating the underestimated support for repeal and nearly every beliefs and opinions of others (Epley et al. 2004; Republicanstafferover-estimatedsupportforrepeal.On Nickerson 1999). We found the clearest case for ego- both the minimum wage and infrastructure spending, centric bias in health policy. On average, staffers who nearly all GOP staffers underestimated their con- supported the ACA repeal overestimated constituent stituents’ support. Democratic staffers tended to underestimate support, on average, though the bias 21 We discuss these explanations separately below, reflecting our view was less pronounced than with their Republican coun- that we cannot make strong causal claims with our observational data. terparts. Thus, overall, we find a conservative bias in Nevertheless, we also provide multivariate regressions adjudicating staffers’ estimations. the importance of each factor in Appendix L.

9 Alexander Hertel-Fernandez, Matto Mildenberger, and Leah C. Stokes

TABLE 1. Correlates of the Staffer Representation Gap

Factor Hypothesis Measure Result preview

Electoral Staffers in offices facing Staffers’ predictions of No evidence supporting competitiveness more competitive races 2016 race margin; this hypothesis (Figure 5) will more accurately per actual margin of victory ceive constituent (Appendix H) preferences Staffer preferences Staffers will offer more acc Staffers’ personal support Staffers’ own beliefs are urate estimates of their for policies correlated with constituents’ preferences misperceptions (Figure 6) if they share their constituents’ preferences Staffer experience Staffers with greater work Staffers’ years of service in No evidence supporting experience in Congress Member office and this hypothesis (Appendix I) will be more likely to Congress accurately perceive constituent preferences Interest groups Staffers with greater contact Self-reported reliance on Interest group contact and with corporate groups interest groups (including campaign contributions will less accurately perceive mass-based advocacy are both correlated with constituent preferences; groups, business misperceptions (Figure 7 staffers with greater contact associations, and think and Appendices J and K) with mass-based groups will tanks); campaign more accurately perceive contributions from industry constituent preferences groups to staffers’ Members in the last electoral cycle; local labor union density

support for repeal while staffers who opposed repeal explanation, we compare staffer tenure against staffer- underestimated constituent support for repeal. By constituent preference mismatches. We find no rela- contrast, all staffers underestimated their constituents’ tionship. In none of the policy domains we examined support for climate regulations, infrastructure spend- was staffer experience, either in a Member’s office or in ing, boosting the minimum wage, and gun background Congress in general, related to the accuracy with which checks, although this dynamic was significantly mod- staffers could discern their constituents’ attitudes. erated by staffers’ own beliefs. This finding indicates a Staffers who have served in Congress for longer are not potential role for staffers’ own perceptions and attitudes more likely to get their constituents’ preferences right in accounting for the representation gap between compared to newer staffers (see Appendix I for full Congressional aides and the public. Our evidence results). suggests that, despite any political incentives that may Finally, we examined whether interest group contact exist to reward unbiased estimation of constituent could account for the representation gap. In the absence beliefs, staffers do not transcend common egocentric of strong signals from ordinary citizens, staffers might bias. Divergence in staffer misperceptions among indi- rely on interest groups that claim to represent con- viduals who agree and those who disagree with each stituents within a staffer’s state or district. Staffers might policy statement are statistically significant in all five even simply substitute interest group preferences for policy domains; substantively, the differences range from those of their constituents. Building on the interest 17-percentage points (infrastructure) to 36-percentage group literature (Schlozman et al. 2012; Walker 1991), points (gun regulations). we distinguish between mass-based interest groups and The third factor we considered was staffer work corporate-based interest groups. Business-oriented experience. There was broad variation in the length of groups represent the interests of for-profit companies. time that senior staffers in our survey had worked for Mass-based groups represent individuals from the mass their particular Member, and for Congress in general. public. We expect that greater relative staffer contact On average, staffers reported working for their Member with mass-based groups will help staffers assemble a for about five years, and in Congress for nine. It could be more accurate perception of their constituents’ pref- the case that staffers with more work experience are erences given that mass-based groups have closer more knowledgeable about their constituents’ prefer- interactions with the public. By contrast, we hypothe- ences, either because they have had more time to learn size that greater relative staffer contact with business- the public’s attitudes, or because they gained skills based groups will result in less accurate perceptions of in estimating constituent opinion. To assess this public opinion because business-based groups are less

10 Legislative Staff and Representation in Congress

FIGURE 5. Staffer Misperceptions of Constituent Preferences and Estimated Margin of Victory in Next Election, by Issue Area

likely to represent the preferences of the general rate their reliance on various interest groups on a one public—and may be opposed to them (Gilens and Page through six scale, where six indicated the greatest 2014; Hacker and Pierson 2010). Consistent with the importance.22 We average and sum standardized ver- more general rightward movement of the organized sions of these measures across multiple interest groups business community in recent decades (Hacker and involved in each of the five policy domains.23 Pierson 2016), most business interests in our survey are Our second measure involves the share of a Mem- also aligned with ideologically conservative positions, ber’s campaign contributions in their last election cycle while our mass-based groups are supportive of liberal that came from businesses in specific industries that stands. Regardless of the cause of this division, this related to our policy domains: health providers and complicates efforts to separate the ideological ori- entation of groups from the business versus mass group 22 The survey item was: “We are interested in knowing which groups dimension in our survey. In the discussion that follows, are most important to you when considering legislation. For each we therefore refer to groups by both characteristics. group, please indicate how important the group’s positions, resources, We measure interest group contact in three ways: and information have been to you when deliberating over legislation.” staffers’ self-reported reliance on interest groups rele- 23 We coded Americans for Responsible Solutions, Everytown, the Sierra Club, League of Conservative Voters, and the AFL-CIO as vant to the policy issues we study, data on campaign fl contributions from the major corporate sectors related to liberal, mass-based interest groups. We coded the National Ri e Association, the American Petroleum Institute, the Edison Institute, these policy domains, and an objective measure of one the US Chamber of Commerce, the National Association of Manu- mass group’s strength (labor unions). The self-reported facturers, the Club for Growth, and Americans for Prosperity as measurecomes froma surveyiteminwhich stafferscould conservative or corporate-based interest groups.

11 Alexander Hertel-Fernandez, Matto Mildenberger, and Leah C. Stokes

FIGURE 6. Staffer Misperceptions of Constituent Preferences and Staffer Personal Opinions, by Issue Area. Lines Indicate 95% Confidence Intervals

insurers for the ACA; gun and ammunition manu- congruence between public opinion and policy change facturers for background checks; construction and that“unionsemerged asthe interestgroups withthe most contractors for infrastructure; business associations for consistent and widespread tendency to share the pref- the minimum wage; and extractive industries, including erences of low- and middle-income Americans (Gilens conventional energy power plants, oil, gas, and coal 2012, 161).” And in their analysis of the contribution of extraction, for carbon dioxide emission limits.24 For economic policy to rising top-end inequality, Jacob now, we interpret the correlation between campaign Hacker and Paul Pierson argue that unions are the contributions and staffer perceptions not as signifying “vigorous champion” of “pocketbook issues” for the weight of the contributions themselves, but rather as “middle- and working-class Americans”—not just their the strength of the relationship between a Member’s immediate members (Hacker and Pierson 2010, 143). In office and a particular set of corporate interests. A light of this research, we feel comfortable characterizing greater reliance on contributions from an industrial unions as an especially important mass-based group sector signifies a stronger relationship between an office representing public preferences. and businesses in that sector. The top left panel of Figure 7 summarizes, across The final measure involves the density of labor unions policy domains, the contact that staffers reported with in a staffer’s district or state, which we interpret as a both liberal/mass groups and conservative/corporate measure of union strength. We rely on estimates of labor groups. The horizontal axis is a standardized index of union membership by House district reported in Becher the frequency with which staffers relied on mass groups et al. (2018) and state union density from the Bureau of minus corporate groups. Greater values indicate a Labor Statistics (see Appendix J for more information). greater relative reliance on liberal/mass groups over Unions represent an especially important mass-based conservative/corporate groups. The strong negative group and check on corporate influence. As Kay relationship indicates that staffers who reported Schlozman, Sidney Verba, and Henry Brady explain in relying on liberal/mass over conservative/corporate The Unheavenly Chorus, unions are essentially the only interest groups—at the right end of the x-axis—tended way that nonprofessional, nonmanagerial workers to more accurately perceive their constituents’ pref- receive representation in the organized interest com- erences. By contrast, staffers more reliant on con- munity (see especially chapter 14; Schlozman et al. 2012). servative/corporate groups over liberal/mass-based In a similar vein, Martin Gilens finds in his analysis of the groups—at the left end of the x-axis—were more likely to get their constituents’ opinions wrong. The top right panel reports the campaign contribution measure. We 24 Data from the National Institute on Money in State Politics. We find observe a similar relationship as in the left panel, similar results using logged campaign contributions in dollars as well. indicating that a greater reliance on corporate

12 Legislative Staff and Representation in Congress contributions is correlated with larger constituent revealing they changed their opinion on a policy issue mismatches.25 after receiving a campaign contribution from an The bottom left panel of Figure 7 shows consistent organized interest group. We used a separate experi- results with the other two interest group panels: greater ment to examine how communications from ordinary mass group-based pressure on a Member’s office (in this constituents, citizens’ associations, or employees of a case measured with local union strength) is related to large business influence staffers’ opinions about more accurate perceptions of constituent preferences. pending policy debates and their perceptions of con- Staffers representing districts and states with stronger stituent attitudes. unions were more accurate in their predictions of fi constituent preferences along all ve dimensions. Contribution List Experiment In summary, we find little support for a relationship between electoral pressures or staff tenure and staffers’ Recent experimental research by Kalla and Broockman accuracy in estimating public opinion. We do find a (2015) gives reason to think that interest group electoral relationship between accurate predictions and interest contributions increase the likelihood that high-level group interactions with and donations to offices, as well staffers meet with interest groups. That study and as staffers’ own opinions. Greater contact with liberal, related literature, however, is limited in its ability to mass-based groups is related to more accurate per- gauge whether contributions also shape staffers’ policy ceptions of constituent opinion while greater contact opinions. While contributions facilitate access, those with conservative, corporate-based groups is related to meetings may not necessarily result in staffers changing less accurate perceptions. Staffer contact with these their minds about legislation. For instance, if interest organizations may skew their perceptions of the public groups generally make campaign contributions to allies because the groups are more ideologically conservative who already agree with them, then contributions should than the public along the range of issues we study. That not change staffers’ minds (Hall and Deardorff 2006; mechanism would be very consistent with the analysis in Hall and Wayman 1990). Examining whether staffers Broockman and Skovron (2018), which shows that state change their minds on legislation after their Repre- politicians’ over-estimation of their constituents’ con- sentative or Senator receives campaign contributions servatism can be attributed in part to conservatives’ from an interest group is no easy task, however. It is a greater likelihood of contacting elected officials. mechanism that is difficult to observe without asking Still, these correlations are at best suggestive. And staffers directly. Yet staffers likely face strong incen- importantly, these results cannot explain why it is that tives against responding truthfully to such a question on contact with conservative and corporate groups skews a survey. Accordingly, we use a list experiment to elicit staffer perceptions of the public. To more rigorously test more accurate responses from senior legislative staffers whether interest group interactions can shape staffer about whether electoral contributions from interest perceptions and positions, we turn to evidence from two groups shaped their policy preferences. If interest experiments embedded in the survey. These experi- groups do shape staffers’ policy preferences, this could ments provide more credible causal evidence about the help explain why corporate contributions are correlated role of interest group and donor contact in shaping with a greater staffer-constituent opinion mismatch. staffer perceptions and preferences. In a list experiment, respondents are shown a list of several statements, and then indicate how many items they agree with. Not all respondents see the same list, EXPERIMENTAL EVIDENCE SUPPORTING however. Half of respondents are shown a short list INTEREST GROUPS’ INFLUENCE ON THE without the sensitive item the researcher is interested in REPRESENTATION GAP studying, while the other half sees a list with the sensitive item included. The proportion of respondents agreeing We use two experiments embedded within the staffer with the sensitive statement can be calculated by sub- survey to test whether contact with organized interest tracting the average number of items selected among groupschanges senior legislative staffers’ perceptions of respondents who saw the full list from the average their constituents’ preferences. In a list experiment, we number of items selected by respondents who saw the uncover the proportion of top Congressional aides short list. List experiments have been successfully used in many other contexts to study similar taboo or sen-

25 sitive behaviors (Glynn 2013). We standardized this variable, as with the mass versus corporate In our list experiment, staffers read the following group contact variable, to be comparable across policy domains. “ fi Unfortunately we cannot exactly match the “mass group minus cor- prompt: Below, you will nd a list of [4/5] statements. porate group” measure we used for the self-reported staffer contact Please tell us HOW MANY of them you agree with. We panel because many of the organizations we included in our survey do don’twanttoknowwhich onesyou agreewith,justHOW not make direct campaign contributions to Congressional candidates. MANY. Over the last year, I’ve developed a new per- We can replicate the exact measure for the CO2 limits question, spective about a policy under consideration in Congress however,and we find a verystrongrelationshipbetweenthe balance of after speaking with…” Half the staffers were assigned to contributions from environmental, mass-based groups and extractive the control group, where they saw four items: “a rep- industry, corporate interests and staffer misperceptions. We also calculate business and labor contributions for each staffer’s office and resentative from a business”; “a legislative staffer from again find similar results looking across all issue domains. We include the opposing party”; “a legislative staffer from my own both results in Appendix K. party”; and “a representative from a union.” The other

13 Alexander Hertel-Fernandez, Matto Mildenberger, and Leah C. Stokes

FIGURE 7. Staffer Misperceptions of Constituent Preferences and Interest Group Contact (Self- Reported Reliance on Mass Versus Corporate Groups in Top Left Plot; Corporate Campaign Contributions to Staffer’s Member in Top Right Plot; Labor Union Membership in Bottom Left Plot)

half of staffers were assigned to the treated group, where campaign contributions. Some of these staffers might they saw those same four items plus an additional fifth develop new policy perspectives from campaign con- item: “a group that provided campaign contributions to tributors because they share broad policy objectives. myMember.” Thisisthe sensitiveitemwhose prevalence Others may engage these contributors to secure con- we were attempting to estimate.26 tinued campaign financing. Critically, however, this Staffers in the control condition agreed with an experiment offers evidence that campaign contributors average of 2.84 items (SE 5 0.18), while staffers in the not only shape access to Congressional offices in the treatment condition agreed with an average of 3.30 first place (Hall and Deardorff 2006; Hall and Wayman items (SE 5 0.19). The difference in means between the 1990; Kalla and Broockman 2015), but that contact with conditions is 0.454 (SE 5 0.27). This indicates that 45% these contributors shapes staffer policy perspectives. In of staffers agreed they had developed a new perspective this sense, the list experiment suggests a potential about a policy after speaking with a group that provided mechanism for the correlations between corporate campaign contributions to their Member.27 This result interactions and staffer-constituent preference mis- suggests that a potentially large share of senior Con- matches we uncovered in the observational survey gressional staffers have had their thinking about policy evidence above. shaped by interest groups that provided their Member Communications Survey Experiment 26 See Appendix O for a discussion of our design and wording choices. Our second survey experiment focused on another 27 Two-tailed test: p 5 0.098, one-tailed test: p 5 0.049; clustered by strategy that interest groups use to shape Congressional office. That we find a significant effect at all given such a small sample fi policymaking: having their members write to Congress size is striking; while our con dence intervals are necessarily wide on pending policy proposals. These tactics are well- given this small sample, we also note that, ultimately, it is the direction of the effect that matters most for our argument that interactions with documented in past work on interest group behavior campaign contributors may skew staffer perceptions of public (e.g., Kollman 1998; Schlozman et al. 2012; Walker preferences. 2014). In particular, large companies increasingly use

14 Legislative Staff and Representation in Congress employee correspondence to shape legislative behavior provided on a one through four scale (“very”, “some- on bills relevant to their firms and industries (Hertel- what”, “not very”, and “not at all”), with higher values Fernandez 2018; Walker 2014).28 Legislative staffers’ indicating greater agreement with the question. Each of role in mediating communications to Members is less these three questions taps into a potential mechanism well understood, however. This oversight is important for how interest group-initiated communications with because we know staffers play a central role in the Congress could change how staffers—and ultimately, information chain connecting outside communications Members of Congress—perceive their constituents’ to Members’ actions. opinions and use those perceptions to inform policy Existing literature on Congressional constituent decisions. By randomly assigning staffers to different correspondence is similarly unclear on whether these hypothetical letters, we can estimate the effect of a communications provide an opportunity for all interests staffer receiving messages that were coordinated in to be represented in the policymaking process, or letter-writing drives from large businesses or citizen whether they advantage certain voices over others. In a advocacy groups, or simply written by individual citi- series of correspondence experiments, Butler (2014) zens.29 As with any survey experiment, the treatment shows that municipal politicians discount constituent and related questions are stripped of much of the rich communications from individuals who appear to be of political context that surrounds the actual legislative low socioeconomic status because elected officials process. Yet the virtue of this stripped-down experi- perceive those individuals to be less knowledgeable ment is that any differences we observe in how about the issues described in those communications. staffers report evaluating these constituent communi- This research suggests important ways that politicians cations can be attributed to variation in the treatment might discount public attitudes of certain groups. We conditions—and not other characteristics of the staffers aim to extend this research, examining how political themselves. Given our interest in understanding the officials evaluate messages from individual citizens effect of organized interest groups on staffers’ per- alongside interest groups, and especially business. We ceptions of constituent opinion, we focus our attention are thus interested in knowing whether interest group- on evaluating the effects of these treatments,rather than initiated communications carry especially important the other conditions in the experiment. It is important to weight in the minds of Congressional staff compared note that given three conditions, there are small sample with individuals’ communications. Can interest groups, sizes (N 5 17–36) in each treatment group (moreover, and particularly businesses, use Congressional com- we did not pre-register our hypotheses). munications as a way of disproportionately represent- With these caveats in mind, staffers reported that they ing their interests in the policymaking process, thus would be more likely to mention employee-authored skewing staffer perceptions of constituents? letters to their bosses, but the differences with the To evaluate these questions, we designed an ex- constituent or citizens’ group conditions were not sig- periment wherein senior Congressional staffers were nificant at this sample size. Staffers were also more likely instructed to imagine that they had received constituent to indicate that employee letters would shape their communications on a pending policy issue. Staffers were thinking and recommendations about legislation, but randomly assigned to receive different descriptions of the difference between employee letters and letters the constituent communications, with the variations from ordinary constituents was not significant. When indicated in brackets as follows: “Imagine your office is comparing employee letters against both citizens group considering a bill that is under debate in Congress. Your letters and ordinary constituents taken together, which office receives [2; 20; 200] letters from constituents improves the sample size, there is a larger effect (0.32 [supporting; opposing] this bill. The letters have very units on the one to four scale, p 5 0.089; 16-percentage [similar; different] wording to one another. The letter points on the “very” or “somewhat” important scale, writers identify themselves as [employees of a large p 5 0.17). company based in your constituency; constituents; The picture is more striking when looking at the final members of a nonprofit citizens group].” question we posed to staffers: how representative they We then asked staffers three questions about the believed the letters were of their constituents’ opinions effect of those letters on their deliberations over the as a whole. Employee messages are substantially more legislation in question: “How likely are you to mention relevant as staffers form perceptions of their con- these letters to your Member?”; “How significant would stituents’ opinions than either messages from individual these letters be in your advice to your Member about citizens or members of nonprofit advocacy groups. A their position on the bill?”; and “How representative do full 62% of staffers who saw messages purportedly from you think these letters are of your constituents’ opin- employees of the same large business reported that the ions?” Responses to the three questions were all messages were “very” or “somewhat” representative of

28 See for instance the account provided in Hertel-Fernandez (2018): 29 We recognize the challenges with ensuring that staffers were “A lobbyist for a telecommunications company reported … that responding under conditions of “information equivalence” (Dafoe recruiting workers to write to a member ‘creates a heightened sense of et al. 2017). In response, we note that we have varied a number of importance of an issue’ and permits their lobbying team to bring those important background conditions about the bill and correspondence. contacts up in one-on-one meetings with the member and his or her In addition, we do not think any of these conditions would present the staff. The lobbyists might say, ‘We have 3,500 workers in your district sort of extreme “edge cases” that Dafoe et al. are most concerned and this is an important issue for them’ (pages 50–1).” about in their paper.

15 Alexander Hertel-Fernandez, Matto Mildenberger, and Leah C. Stokes their constituents’ opinions, compared to just 34% of citizens (Gilens 2012; Gilens and Page 2014). Overall, staffers who saw letters from ordinary constituents and these results suggest a significant representation gap for 29% of staffers who saw letters from nonprofit advocacy average American citizens. group members. More dramatically, over a third of We contribute to this research, examining a poten- staffers with letters from nonprofit advocacy groups and tially new mechanism for the breakdown in respon- around 10% of staffers with letters from individual siveness: the staffers serving Members of Congress and constituents reported that those letters were “not at all their contact with organized interest groups, partic- representative” of their constituents’ real opinions. The ularly businesses and ideologically conservative groups. difference between the employee letter condition and Through our observational analysis and survey the ordinary constituent condition was quite large, experiment on constituent communications, we have representing 0.39 units on the one to four scale (p 5 presented evidence consistent with the argument that 0.019), as was the difference between the employee when senior Congressional staff come into contact with letter condition and the citizens group condition (0.68 corporate or conservative interests, they are less likely units, p , 0.01).30 to have an accurate picture of what their constituents Taken together, these results provide further evi- want. dence for a mechanism that businesses can use to Our results cast a rather pessimistic light on dem- shape Congressional policymaking (Hertel-Fernandez ocratic representation in Congress. Although senior 2018;Walker2014). They also illuminate a potential staffers responsible for advising Representatives and mechanism by which staffers’ perceptions of their Senators overwhelmingly report they would like to constituents’ preferences can be skewed away from the base their decisions and recommendations on con- attitudes held by most citizens. If a slew of employee stituent opinion, in practice these staffers have only a letters arrive supporting a particular bill, staffers may limited understanding of constituent preferences be more likely to think that their district or state as a across important policy issues. Our results also suggest whole supports that bill, even if those employee letters that far from being a tool for democratizing access to are not representative of public opinion. The results lawmakers, in practice Congressional correspondence from the staffer communications survey, then, com- can be a mechanism through which organized business plement our observational analysis of staffer percep- disproportionately influences the policymaking proc- tions of district and state opinion, showing one ess at the expense of other interests (cf. Yackee and potential mechanism by which organized interest Yackee 2006). groups—and especially businesses—can distort the The staffer survey we used to reach these conclusions representation of constituents’ opinions in the eyes of afforded us a number of advantages. It allowed us to Congressional staff. study the top aides responsible for advising Members of Congress on legislative and policy decisions. We were also able to combine both observational and exper- fi LEGISLATIVE STAFFERS, ORGANIZED imental evidence, nding support for interest groups and egocentric bias in driving the mismatch. Yet our INTEREST GROUPS, AND AMERICAN fi DEMOCRACY ndings also have several important limitations. We can only speak to staffer representation on the issues that we As economic inequalities expand in the United States, included on the survey. It may be that there are issues on many scholars and citizens have expressed concern that which staffers are much better positioned to evaluate those at the top of the income distribution are advan- their constituents’ preferences. Still, given the paucity of taged in the political process. A growing body of representation on these five issues, which span very research suggests that these fears may be well-founded. different types of highly salient policies during an Wealthy individuals and business interests are more election year, we think that this is unlikely to be true. involved in the political process than groups repre- Our findings about representation are also subject to senting the economically disadvantaged or workers broader concerns about the measurement of mass (Drutman 2015; Schlozman et al. 2012). Moreover, opinion. Does it matter if legislative staffers do not when researchers compare policy changes to public correctly perceive their constituents’ opinions if con- preferences, results show business’ and economic elites’ stituent opinions simply represent ill-informed “non- preferences are much more predictive of policy change attitudes” (Converse 1977)? We are agnostic in this than are the preferences of lower- and middle-income analysis about the quality of constituent attitudes themselves, simply noting that we have chosen salient policy issues that Congress has recently considered. 30 Although we are hesitant to conduct subgroup analyses with this These are the kinds of issues where we should expect experiment given our small sample, in Appendix P we show that citizens to have developed relatively clearer attitudes. staffers from districts and states with higher levels of unemployment Moreover, even giving staffers a generous benefit of the were much more likely to interpret employeecorrespondence as being doubt in their precision, the magnitude of the mismatch more representative than were staffers from areas of lower unem- between staffer perceptions and constituent prefer- ployment. The fact that staffers who were potentially much more ences is quite striking. attuned to unemployment in their constituency were much more responsive to the employee letters is consistent with theories of A third limitation is that our results are necessarily structural business power. At the same time, we reiterate that we view bounded to the contemporary Congress. Recent these results as requiring follow-up research. increases in Congressional contact under the Trump

16 Legislative Staff and Representation in Congress

Administration may change some of these findings. It is Broockman, D., and C. Skovron. 2018. “Bias in Perceptions of Public also an open question whether these results would hold Opinion Among Political elites.” American Political Science Review. AU4 in other times or across different legislative bodies. Butler, D. 2014. Representing the Advantaged. New York: Cambridge Future work assessing legislative staffers’ role in University Press. mediating mass responsiveness in the states, or in other Carnes, N. 2013. White-Collar Government. Chicago: University of countries, would be very useful. Chicago Press. Additional research could also examine the mecha- Carroll, R., J. Lewis, J. Lo, N. McCarty, K. Poole, and H. Rosenthal. fi 2015. “‘Common Space’ DW-Nominate Scores with Bootstrapped nisms underpinning our ndings in greater detail. In Standard Errors.” www.voteview.com/dwnomin_joint_house_and_ particular, future work ought to disentangle con- senate.htm. servative and business-group contact with staffers. Do CMF. 2011. “Communicating with Congress.” Report, Congressional both types of groups skew staffer perceptions of the Management Foundation. public? Our findings in the communications experiment Converse, P. 1977. “The Nature of Belief Systems in Mass Publics.” In Ideology and Discontent, ed. D. Apter. New York, NY: Free Press. offer some suggestive evidence that business groups Cook, P., and K. Goss. 2014. The Gun Debate. Oxford University may independently shape staffer perceptions of their Press. AU5 constituents. But additional work might examine the Curry, J. 2015. Legislating in the Dark. Chicago: University of Chicago reasons behind this bias in more detail. For example, Press. Dafoe, A., B. Zhang, and D. Caughey. 2017. Information Equivalence staffers may be more deferential to business repre- in Survey Experiments. Unpublished manuscript. AU6 sentatives because they view employees as more DeGregorio, C. 1988. “Professionals in the U.S. Congress: An knowledgeable about policy issues than ordinary citi- Analysis of Working Styles.” Legislative Studies Quarterly 13 (4): zens, especially on issues that affect their firms. 459–76. Ultimately, our analysis should not be seen as an Druckman, J., and L. Jacobs. 2015. Who Governs? Chicago: Uni- versity of Chicago Press. indictment of staffers alone. If the public does not con- Drutman, L. 2015. The Business of America is Lobbying. New York, sistently communicate its preferences to Congress—and NY: Oxford University Press. interest groups do—then it is hard to expect staffers to Enns, P. 2015. “Relative Policy Support and Coincidental Repre- accurately perceive what the public wants. In addition, sentation.” Perspectives on Politics 13 (4): 1053–64. some citizens may be more likely to contact their rep- Epley, N., B. Keysar, L. Van Boven, and T. Gilovich. 2004. “Per- spective Taking as Egocentric Anchoring and Adjustment.” resentatives than others, whether these disparities occur Journal of Personality and Social Psychology 87 (3): 327. by issue, socioeconomicbackground, or geography. Such Erikson, R. 2015. “Income Inequality and Policy Responsiveness.” asymmetric mobilization can lead to distorted signaling, Annual Review of Political Science 18: 11–29. making it harder for staff to understand the preferences Erikson, R., N. Luttbeg, and W. Holloway. 1975. “Knowing One’s of the whole public they seek to serve (Stokes 2016). District: How Legislators Predict Referendum Voting.” American Journal of Political Science 19 (2): 231–46. Though efforts to increase mass participation in the Fenno, Jr. R. 1978. Home Style. New York: HarperCollins. policymaking process are likely to be challenging, they Free, L., and H. Cantril. 1967. The Political Beliefs of Americans. New could help to close the staffer-constituent opinion gap Brunswick, NJ: Rutgers Press. and ensure better representation in Congress. Regard- Gilens, M. 2012. Affluence and Influence. Princeton: Princeton Uni- versity Press. less of how future scholars and activists choose to pursue Gilens, M., and B. Page. 2014. “Testing Theories of American Politics: the research questions and normative issues raised in this Elites, Interest Groups, and Average Citizens.” Perspectives on paper, the bottom line is that legislative staffers—and Politics 12 (3): 564–81. their resources, perceptions, and interactions—merit far Glynn, A. 2013. “What can We Learn with Statistical Truth Serum? more attention in political science. Design and Analysis of the List Experiment.” Public Opinion Quarterly 77 (S1): 159–72. Hacker, J., and P. Pierson. 2010. Winner-Take-All Politics. New York, NY: Simon and Schuster. SUPPLEMENTARY MATERIAL Hacker, J., and P. Pierson. 2016. American Amnesia. New York, NY: Simon and Schuster. To view supplementary material for this article, please Hall, R. 1996. Participation in Congress. New Haven: Yale University Press. visit https://doi.org/10.1017/S0003055418000606. Hall, R., and A. Deardorff. 2006. “Lobbying as Legislative Subsidy.” Replication materials can be found on Dataverse at: American Political Science Review 100 (1): 69–84. https://doi.org/10.7910/DVN/OWQNVF. Hall, R., and F. Wayman. 1990. “Buying Time: Moneyed Interests and the Mobilization of Bias in Congressional Committees.” American Political Science Review 84 (3): 797–820. Hammond, S. W. 1996. “Recent Research on Legislative Staffs.” Legislative Studies Quarterly 21 (4): 543–76. REFERENCES Hedlund, R., and P. Friesema. 1972. “Representatives’ Perceptions of Constituency Opinion.” Journal of Politics 34 (3): 730–52. Arnold, R. D. 1990. The Logic of Congressional Action. New Haven: Hertel-Fernandez, A. 2018. Politics at Work: How Companies Turn Yale University Press. Their Workers into Lobbyists. New York, NY: Oxford University Bartels,L.2008. Unequal Democracy. Princeton:Princeton University Press. Press. Howe, P., M. Mildenberger, J. Marlon, and A. Leiserowitz. 2015. Becher, M., K. Kappner, and D. Stegmueller. 2018. “Local Union “Geographic Variation in Opinions On Climate Change at State Organization and Lawmaking in the U.S. Congress.” Journal of and Local Scales in the USA.” Nature Climate Change 5 (6): AU3 Politics 80 (2). 596–603. Bertrand, M., M. Bombardini, and F. Trebbi. 2014. “Is It Whom You Jacobs, L., and R. Shapiro. 2000. Politicians Don’t Pander. Chicago: Know or What You Know? An Empirical Assessment of the The University of Chicago Press. Lobbying Process.” American Economic Review 104 (12): Jacobs, L., and T. Skocpol. 2015. Health Care Reform and American 3885–920. Politics. New York, NY: Oxford University Press.

17 Alexander Hertel-Fernandez, Matto Mildenberger, and Leah C. Stokes

Kalla, J., and D. Broockman. 2015. “Campaign Contributions Facil- Montgomery, J., and B. Nyhan. 2016. “The Effects of Congressional itate Access to Congressional Officials: A Randomized Field Staff Networks in the U.S. House of Representatives.” AU7 Experiment.” American Journal of Political Science 60 (3): 545–58. Nickerson, R. 1999. “How We Know—and Sometimes Misjudge— Kingdon, J. 1984. Agendas, Alternatives, and Public Policies. Boston: What Others Know: Imputing One’s Own Knowledge to Others.” Little, Brown. Psychological Bulletin 125 (6): 737. Kingdon, J. 1989. Congressmen’s Voting Decisions. Ann Arbor: The Romzek, B., and J. Utter. 1997. “Congressional Legislative Staff: University of Michigan Press. Political Professionals or Clerks?” American Journal of Political Kollman, K. 1998. Outside Lobbying. Princeton: Princeton University Science 41 (4): 1251–79. Press. “ Malbin, M. 1980. Unelected Representatives. New York: Basic Books. Salisbury, R., and K. Shepsle. 1981. U.S. Congressman as Enter- Mann, T., and N. Ornstein. 2016. It’s Even Worse than it Looks. New prise.” Legislative Studies Quarterly 6 (4): 559–76. York, NY: Basic Books. Schlozman, K. L., S. Verba, and H. Brady. 2012. The Unheavenly Mayhew, D. 2004. Congress: The Electoral Connection, 2nd edition. Chorus. Princeton: Princeton University Press. New Haven: Yale University Press. Stokes, L.. 2016. “Electoral Backlash against Climate Policy: A McCrain, J. (forthcoming). “Revolving Door Lobbyists and the Value Natural Experiment on Retrospective Voting and Local Resistance of Congressional Staff Connections.” Journal of Politics. to Public Policy.” American Journal of Political Science 60 (4): Mildenberger, M. 2015. “Fiddling while the World Burns: The Logic of 958–74. Double Representation in Comparative Climate Policymaking.” Swers, M. 2002. The Difference Women Make. Chicago: University of PhD thesis. New Haven: Yale University. Chicago Press. Mildenberger, M., J. R. Marlon, P. D. Howe, and A. Leiserowitz. 2017. Walker, E. 2014. Grassroots for Hire. New York, NY: Cambridge “The Spatial Distribution of Republican and Democratic Climate University Press. Opinions at State and Local Scales.” Climatic Change 145 (3–4): Walker, J. 1991. Mobilizing Interest Groups in America. Ann Arbor, – 539 48. MI: University of Michigan Press. Constituency Representation in Congress Miler, K. 2010. . New York, Yackee, J. W. and S. W. Yackee. 2006. “A Bias towards Business? NY: Cambridge University Press. fl ” Miller, W., and D. Stokes. 1963. “Constituency Influence in Con- Assessing Interest Group In uence on the U.S. Bureaucracy. gress.” American Political Science Review 57 (1): 45–56. Journal of Politics 68 (1): 128–39.

18